Regarding the issue of memory

[Nora-jsu]

Hi, I am trying to replicate your work, and I see that you trained using V100. I would like to know how many V100 GPUs you used for parallel training and how much GPU memory was occupied during the training